Re: [RFC PATCH 3/3] x86/boot: Use alternatives based selector for 5-level paging constants

[Ingo Molnar <mingo@kernel.org>]

* Ard Biesheuvel <ardb@kernel.org> wrote:

> > All of those change depending on
> > whether we do 5-level page tables or not, so the whole argument that
> > "pgtable_l5_enabled() is special" is just wrong to begin with.
> >
> 
> In my original patch, which is the one Ingo objected to, 
> pgtable_l5_enabled() is unambiguously based on whether CR4.LA57 is 
> set.

So I didn't really object to the simplification aspect - I was 
criticizing the current state of LA57 handling, regardless of your 
patch. In fact in that thread I supported the simplification aspect:

  > > Anyway, I'm not against Ard's simplification patch as a first step, and
  > > any optimizations can be layered on top of that.

But in hindsight I can see how my first reply came away as 
disagreement...

> In the light of the above, care to comment on the previous approach?
> 
> https://lore.kernel.org/all/20250504095230.2932860-28-ardb+git@google.com/
> 
> That also uses the ALTERNATIVE_TERNARY() so the CR4 access gets
> patched away, and I'm happy to take suggestions how to improve that.

I still think we should introduce a LA57_ENABLED synthethic cpufeature 
flag or so for the MM constants and all the late facilities, and go 
from there.

Thanks,

	Ingo